Investigation Of Fabrication And Defects Of SiC Single Crystal

Silicon carbide is a promising semiconductor material to realize high-power, high-frequency semiconductor devices, and service in high-temperature, corrosive, and high-radiation environments, due to its wide bandgap, high-breakdown electric field strength, high-saturated drift velocity and chemical stability.Recently, SiC single crystal growth is a research focus in semiconductor materials field. In all technical problems of SiC single crystal growth, the formation and avoiding of micropipes is the most important one. Carbon is an important source material for SiC single crystal growth, for it has significant effects on crystal growth rate and formation of the defects. Nucleation of SiC polycrystal possibly occurs on the lid where seed is stuck, and those polycrystal restrain the radial growth and enlargement of seed. All previous kinetic simulations were based on transient models, in which the changes of growth characters were not considered. So those models cannot reflect the actual growth process. Aiming at the problems above, our research investigated the growth process and defects of crystal roundly. Some valuable results for practical application and theoretical study were obtained.The model of stacking fault with a core of inclusion for micropipe formation was proposed on the base of the investigation on shapes, sizes, distribution, origin and growth of micropipes. Using this model, the shape, size and distribution of micropipe can be successfully explained, and consequently the methods to reduce the micropipes are proposed. Si inclusions can be restrained by adding active carbon on the top of SiC source powder, which decreases the density of micropipes. Using ( 11 20) and ( 1 100) crystal faces as the growth surfaces, the generation of screw dislocation and micropipes was restricted consumedly.The function of carbon (or graphite) on SiC single crystal growth process was investigated roundly. Graphite crucible is an important carbon source for the SiC crystal growth through the formation of SiC2 by the reactions between crucible wall and Si. High graphitization degree or increase of graphitization degree for crucible during growth process will degrade the growth rate of crystal and augment Si inclusions and defects relative to them. Appending active carbon could provide plenty of carbon source for crystal growth and enhance the growth rate of crystal. Furthermore, it would restrain the formation of Si liquid phase, which reduces the occurrence of defects. The radial growth rates and shapes of crystals are related to active carbon. The radial growth rate of crystal grown with active carbon increases gradually from side to center of crystal, and the surface of crystal shows convexity. However, the radial growth rate of crystal grown without active carbon decreases gradually from side to center of crystal, and the surface of crystal shows concave. The crystal growth rate was also affected by the position of active carbon in the source powder. When active carbon was put under the source powder, it cannot provide carbon source for crystal growth process. When active carbon was put in the middle, it can only supply carbon source to the source powder below. Active carbon can provide carbon source from the beginning to the end of crystal growth process when it was put on the top. The effects of the lid material on nucleation of SiC polycrystal were investigated, and the results indicated that SiC polycrystal can form quickly in the inceptive stage of the growth process, which restricts the radial direction growth of seed. Ta lid can restrain the nucleation of SiC polycrystal in the inceptive stage of the growth process, which creates better condition for seed growth along the radial direction.In numerical simulations, by renewedly defining the mass transport distance, a dynamic kinetics model is set up. According to dynamic model, the growth rate decreases gradually with the increasing of growth time. There will be an 8.2% error between the growth rate of transient and dynamic models when growth time is 5 hours. This error is not too large, so the transient model can be used to compare with experiments. However, there will be a large error of 38.7% between two models when growth time is 27 hours. In this case, the transient model will not reflect the real experiments.